TW479413B - D/A converter, its method, and the data interpolation device - Google Patents

Abstract

The present invention discloses a D/A converter, its method, and the data interpolation device, which is configured with: a digital waveform generation portion 10, for synthesizing the digital data of the basic waveform generated from the discrete data corresponding to the input by the iterative operation; a revolution operation portion 20, for over-sampling on the output from the digital waveform generation portion 10 and conducting iterative operation; and, a D/A conversion portion 30, for D/A conversion on the output from the revolution operation portion 20 and conducting amplitude modulation on the discrete data with the basic digital waveform and synthesizing by the iterative operation. Then, the method can obtain the sequential interpolation value with only the over-sampling and the iterative operation so that it can eliminate the need for configuring with a low-pass filter and the deterioration of the phase characteristic caused by the filter.

Description

479413

Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the Invention (1) The present invention relates to a digital-to-analog converter (D / A converter) that can convert decentralized digital data into continuous analog signals and Methods and data interpolation devices are particularly suitable for use in D / A conversion of speech data. In recent digital audio devices, such as a CD player or a DVD player, in order to obtain a continuous analog voice signal from decentralized digital voice data, a D / A converter using oversampling technology is used. In order to interpolate data between distributed input digital data, this D / A converter uses a digital filter to increase the sampling frequency in a virtual way. Generally, a digital filter is used to maintain the interpolation in the sample holding circuit to generate a stepped signal. After the waveform, it is passed through a low-pass filter to output a smooth analog voice signal. The data interpolation usually performed by the digital filter included in the D / A converter uses a specimenization function called the s i n c function. Figure 1 is an explanatory diagram of the s i n c function. The s i n c function appears when inverse Fourier transform of the 5 function of Dirac. When the specimen frequency is f, it is defined by s 1 η (π f t) / (π f t). This s i n c function is only 1 at the specimen point at t = 0, and 値 at all other specimen points. Fig. 2 is a diagram showing the relationship between the scattered data and the interpolated frame therebetween. For example, the speech signal with a smooth analogy is sampled at a certain time interval, and it is quantized to obtain the dispersed speech data as the sample data. The D / A converter inputs this dispersed digital voice data 'and outputs an analog voice signal during which the continuity of the connection is processed by the above-mentioned interpolation of the s i n c function. In Figure 2, it is assumed that the specimen points t 1, t 2, t 3, and t ζ are equally spaced. The paper size applies the Chinese National Standard (CNS) A4 (210 X 297 mm) -4-ϋ an n ϋ ί βϋ ϋ n ί I · tm inn MM§ ϋ ϋ ϊ, a an «ϋ ϋ« ϋ nn an 1 1 (Please read the notes on the back before filling out this page) 479413 A7 B7 V. Points of the invention description (2) The scattered data is Y (tl), Y (t2), Y (t3), Y (t4) 'If it is required, for example, a certain position t 0 between the specimen points t 2 and t 3 ((Please first Read the notes on the back and fill in this page) Interpolation 値 Υ at a distance from t 2) ° Generally speaking, when using interpolation function to find the interpolation 値 y, you can first obtain the scattered data given値 'of the specimenization function at the interpolation position t 0, and then use this to perform the folding operation. Specifically, according to each specimen point of t1 to t4, the height of the beetle at the center position of the specimenization function is consistent. Add them all up. Furthermore, as time passes, the interpolation position t 〇 will move 'but the position corresponding to each specimen position will also change as time passes', so the interpolation 値 y (t 0) also changes continuously. Obtain a continuous analog signal that smoothly connects scattered data. However, the conventional D / A converter using the above-mentioned oversampling technique is because the stepped signal waveform obtained by interpolation is passed through a low-pass filter to generate a smooth analog signal, so there is an analog output The signal has a problem that its phase characteristics are deteriorated by a low-pass filter. Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs. At the same time, the above sine function is a function that converges to ± 0. Therefore, when the correct interpolation is required, one must find the sinc function of all the scattered data, and then Add up. In practice, however, due to the relationship of processing capacity or circuit size, the range of scattered data to be considered is limited for digital filter processing. Therefore, the interpolation interpolation obtained contains errors in removing the mantissa, and there is a problem that a correct interpolation interpolation cannot be obtained. In this way, the traditional D / Α converter in the application of the oversampling technology 'because -5- this paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 479413 Α7 Β7 V. Description of the invention (3) ( Please read the precautions on the back before filling this page.) The phase characteristics are degraded by the low-pass filter. At the same time, the digital filter using the sine function is used to remove the mantissa error. Therefore, there is a problem that the output waveforms corresponding to these are distorted and the quality of the output voice is deteriorated. The present invention has been made in order to solve this problem, and its purpose is to obtain an output waveform with little distortion so that it can improve the quality of output speech. The digital-analog converter of the present invention is obtained by supersampling and more than one moving average operation or folding operation to synthesize the digital data of the basic waveforms corresponding to the input n scattered data to each other, thereby obtaining After interpolating the digits of the above dispersed data, the digital data including the interpolated digits is transformed into an analog quantity. In other forms of the present invention, the digital data corresponding to the basic waveforms of the input n scattered data are synthesized with each other by oversampling and moving average calculation or folding operation, and the synthesized digital data is then moved. After averaging or folding calculation to obtain the interpolated digits of the above-mentioned scattered data, the digital data including the interpolated digits are converted into analog quantities. The form includes: a synthetic component that synthesizes the digital data corresponding to the basic waveforms of the input n scattered data with each other through a moving average operation or a folding operation; it is carried out in a plurality of segments, and For digital data, the input data is sampled at twice the frequency of the previous paragraph, the data obtained is added, and each of these data is shifted from a certain phase, and they are added to the output of the next stage. Sampling component; advance all the data obtained through the above-mentioned supersampling component-6-This paper size applies Chinese national standards CNS) A4 specification (210 X 297 mm) 479413 A7 B7 V. Description of the invention (4) Operation components for moving average calculation or folding operation; and Transforming each data obtained by the above operation components into D of analog quantity / A transform component. (Please read the precautions on the back before filling this page.) Other forms of the present invention include: digital data corresponding to the basic waveform of one of the n dispersed data input according to the reference frequency clock pulse, staggering the reference frequency clock with each other. The pulses are divided and added to synthesize the components for data synthesis; in a range of multiple segments, the digital data generated by the above-mentioned synthesis components are sampled at the frequency of twice the previous segment, and each of the data obtained is sampled. The data 値 and the data 値 which are staggered by half the clock pulse are added separately and output to the next-stage processing oversampling component; for each data 藉 obtained by the above-mentioned oversampling component, the above-mentioned oversampling Under the frequency clock pulse of the last segment of the component, each data is staggered by one clock pulse and added to perform a moving average operation or a folding operation; and each data obtained by the above operation component is converted into an analogy. Quantitative D / Α conversion component. Printed here by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs, the above-mentioned composite component is provided with, for example, decentralized data that can be sequentially input according to the reference frequency clock pulses and sequentially delay n reference components of the reference frequency clock pulses; and Each data 输出 output from the n delay members is multiplied by each gain 对应 corresponding to the basic digital waveform. At the same time, the multiplication results are added and output to the multiply-add member of the supersampling member. At the same time, the above-mentioned supersampling component includes: for example, each piece of data of the digital data generated by the above-mentioned synthesis component, sampling with a clock pulse that is twice the frequency of the above-mentioned reference frequency, separately sampling each of the obtained data, and staggering this. Each piece of data divided by half a clock pulse is added as the first calculation component; the Chinese National Standard (CNS) A4 specification (210 X 297 mm) is applied to this paper size. Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs and Consumer Cooperatives 479413 A7 B7____ 5 5. Description of the invention (5) Each data obtained from the above-mentioned first operation means is sampled with clock pulses of 4 times the above reference frequency, and each of the obtained data is separated by a half clock pulse. A second operation means for adding the data; and each data obtained from the above 2. operation means is sampled with a clock pulse of 8 times the frequency of the reference frequency, and each of the data obtained is This is a third arithmetic component that is staggered by adding data of half a clock pulse. At the same time, the operation means includes, for example, a plurality of delay means for sequentially delaying the majority of the frequency clock pulses of the last stage of the oversampling means, respectively, of the digital data obtained by the oversampling means; and The output component is added and the output component is added. At the same time, the digital-analog conversion method of the present invention includes: by oversampling and more than one moving average operation or folding operation, the digital data corresponding to the basic waveforms of the input n scattered data are synthesized with each other, thereby obtaining A step of calculating the digital interpolation of the scattered data is performed; and a step of D / A conversion of converting each of the digital data including the interpolation of the data obtained by the above calculation into an analog quantity. Other forms of the present invention include: a synthetic step of synthesizing digital data corresponding to the basic waveforms of the input n scattered data by using moving average calculation or folding operation; and associatively moving the synthesized digital data An oversampling step of the oversampling of the averaging operation or the folding operation; a moving average operation or a folding operation on the data obtained by the above-mentioned oversampling to obtain a digital interpolation step of the scattered data; and, The digital data including the interpolated 求 obtained through the above calculation will be applied to the Chinese standard (CNS) A4 specification (210 X 297 mm). (Please read the precautions on the back before filling this page)

479413 A7 B7_ V. Description of the invention (6) D / A conversion step of 値 conversion into analog quantity. (Please read the notes on the back before filling this page) At the same time, the interpolation device of the present invention is based on the basics of the II input of scattered data through oversampling and more than one moving average operation or folding operation. The digital data of the waveforms are synthesized with each other, thereby obtaining the interpolation of the digital data of the above-mentioned scattered data. In other forms of the present invention, digital data corresponding to the basic waveforms of the input n scattered data is synthesized by moving average calculation or folding operation, and the synthesized digital data is oversampled. The obtained data is then subjected to a moving average operation or a folding operation to obtain the interpolation of the digits of the above-mentioned scattered data. When the consumer cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs prints the present invention constituted as described above, the digital data corresponding to the basic waveforms of the input dispersed data are synthesized with each other only by oversampling and moving average calculation or folding operation. Processing, we can obtain the continuous interpolation of the original scattered data, so the result of the D / Α transformation becomes a smooth analog signal. Therefore, it is not necessary to provide a sample holding circuit or a low-pass filter like the conventional one, and it is possible to suppress the deterioration of the phase characteristics caused by the filter. In addition, the function generated from the basic digital waveform of the present invention is a specimenized function of a finite station, so the number of scattered data required to obtain an interpolation frame can be reduced, and when the number of scattered data that should be processed is reduced, Can make it not produce the error that removes the mantissa. Therefore, distortion of the output waveform can be suppressed to a minimum. According to the above points, the quality of the analog signal of the output can be greatly improved. The D / A converter in this embodiment does not use a digital filter for ultra-paper size. Applicable to China National Standard (CNS) A4 (210 X 297 mm) 479413 Λ7 B7 V. Description of the invention (7) After sampling, pass the sample The holding circuit and low-pass filter generate analog signals. The characteristic is that the scattered data that should be input synthesize the digital data corresponding to the basic waveform of the specimen function, and perform supersampling and moving average operations on the obtained data. Or a folding operation (hereinafter referred to as a convolution operation), whereby each interpolation 値 is digitally obtained, and an analog signal corresponding to this is generated. The D / A converter of this embodiment will be described in detail with reference to the drawings. Figures 3 and 4 are diagrams showing the structure of the D / A converter of this embodiment, and Figures 5 to 7 are diagrams for explaining the principle of the D / A converter of this embodiment. First, the principle of the D / A converter will be described with reference to FIGS. 5 to 7 described above. Fig. 5 is an explanatory diagram of a basic digital waveform used in this embodiment. The basic digital waveform shown in Figure 5 is the basis of the specimenization function used in data interpolation by oversampling. This basic digital waveform is created by changing the data to -1, 1, 8, 8, 1, and -1 every clock pulse (CK) of the reference pulse. Here, in order to explain the basic principle of the D / A conversion operation of this embodiment, consider a case where the following processing is applied to the basic digital waveform of FIG. 5. First, the digital data 基本 of the basic waveform shown in Figure 5 is sampled with a clock pulse (2CK) of twice the frequency, and each of the obtained samples 値 is staggered by a half clock of 2CK. The pulses (half-phase) of each sample are summed, thereby performing digitally oversampling with twice the convolution operation with two stages. Next, I will use the two data obtained from the first oversampling, and then two: --- ^ -------- install --- (please read the precautions on the back before filling this page) · Printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs. The paper size applies to the Chinese National Standard (CNS) A4 (210 x 297 mm) -10-Printed by the Employees’ Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs π 479413 A7 B7___ Note (8) Sampling of clock pulses (4CK) with frequency multiples, and add each sample obtained, and add each sample divided by half clock pulses (half phase) staggered by 4CK, and add Therefore, twice the oversampling with the two-stage convolution operation is performed similarly to the first time. Then, the data 値 obtained from the second oversampling is sampled with two clock pulses (8CK) at a frequency twice, and each of the obtained samples 値 is staggered by a half of 8CK. The clock pulses (half-phase) are added to each sample, thereby performing one more double oversampling with two convolution operations. In this way, twice the oversampling and the two-stage convolution operation are performed three times, and then each data obtained by the third convolution operation is performed, and each clock pulse (8CK) with the same frequency as the third is used to The sampling frames are staggered by one clock pulse, and a convolution operation of 8 segments is performed. Fig. 6 is a diagram showing the results of performing the above-mentioned oversampling and convolution operation on the basic digital waveform of Fig. 5. Among them, Fig. 6 (A) shows the results of the first oversampling and convolution operation. In Figure 6 (A), the number list in the first row does not show the results of the double oversampling on the basic digital waveform data in Figure 5. The number in the second row indicates that The sample is staggered by the results of the half phase division. The sequence of the third row indicates the result of adding each sample 算 in the first row and each sample 値 in the second row between the corresponding columns. Figure 6 (B) shows the results of the second oversampling and convolution operation. In FIG. 6 (B), the first row of the sequence indicates that the data shown in the third row of FIG. 6 (A) obtained by the first oversampling and convolution operation is twice oversampled. As a result, the sequence of the second line indicates that the paper size of the first line applies the Chinese National Standard (CNS) A4 specification (210 x 297 public love) (Please read the precautions on the back before filling this page). 479413 Λ7 ___B7____ V. Description of the invention (9) ·, ------------- install --- (Please read the precautions on the back before filling in this page) The results of each sample are staggered. In addition, the sequence of the third row indicates the result of adding each sample 値 in the first row and each sample 行 in the second row between the corresponding columns. Figure 6 (C) shows the results of the third oversampling and convolution operation. In Figure 6 (C), the first row of the sequence indicates that the data shown in the third row of Figure 6 (B) obtained by the second oversampling and convolution operation is twice oversampled. As a result, the sequence in the second row shows the results obtained by staggering the samples in the first row by a half phase. In addition, the sequence of the third row indicates the result of adding each sample 値 in the first row and each sample 行 in the second row between the corresponding columns. Furthermore, due to the limitation on the drawing, a series of series is represented by a two-segment structure. Fig. 6 (D) shows the result of performing 8-stage convolution operation. In Figure 6 (D), the first row of the sequence shows the data shown in the third row of Figure 6 (C) obtained by the third oversampling and convolution operation. The sequence shows the results of sequentially shifting the samples in the first row by a half phase. The sequence of the 9th row shows the result of adding each sample 算 in the 1st to 8th row between the corresponding columns. In addition, here is also a series of sequences represented by a two-segment structure due to limitations in the drawing. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs β When the final sample obtained in line 9 of Figure 6 (D) is D / A transformed, the waveform function shown in Figure 7 can be obtained. signal. The function shown in Figure 7 is that the entire field can be differentiated once. The specimen position t along the horizontal axis has a finite value other than 0 when the sample position t is between 1 and 65. In the remaining fields, all of them are 0. function. In addition, the function has a limit other than 0 in local areas. 'The national paper (CNS) A4 specification (210 X 297 Public Love 1 " 12' " " " " " " " " " " " " Printed by the Consumer Cooperative 479413 B7____ V. Description of the invention (10) When other fields are 0, it is called "limited station". The function in Figure 7 is the maximum value only at the specimen point at t = 3 3, and at t = The four specimen points of l, 17, 49, and 65 become the specimenization function with 0 as its characteristic, and all the sample points required to obtain a smooth analog waveform signal are obtained. Thus, the function shown in FIG. 7 is the specimenization function. , Is a function that can be differentiated once in the entire field, and converges to a finite station of 0 at the specimen position t = 1, 65. Therefore, the specimen function shown in Fig. 7 is used instead of the traditional sine function shown in Fig. 1 to perform Based on the overlap of the scattered data, the difference between the scattered data can be interpolated using a function that can be differentiated once. However, in this embodiment, the overlap (synthesis) of the scattered data is not based on the basic digital waveform of Figure 5. Find Figure 7 The function is executed after the specimen is converted, but as described later in Figures 3 and 4, it is performed digitally before the above-mentioned oversampling and convolution operations. Therefore, only the overlapping digits of the scattered data are implemented. The data can be obtained immediately by performing the above-mentioned oversampling and convolution operation, and the result is equivalent to the specimenization function corresponding to the size of each scattered data in the overlap of Figure 2. The sine function traditionally used is at the specimen point of ± 〇〇 A function that converges to 0. Therefore, in order to obtain the correct interpolation 値, it is necessary to correspond to the scattered data of soil 00, calculate the 値 of the sinc function at the interpolation position, and use these 値 to perform the folding operation. In this regard, this implementation The specimenization function of Fig. 7 used in the morphology converges to 0 at the specimen position t = 1, 6 5. Therefore, it is only necessary to consider the scattered data in the range of t = 1 to 65. Therefore, to obtain a certain interpolation In the meantime, as long as a limited number of dispersions are considered, the paper size applies the Chinese National Standard (CNS) A4 specification (210 X 2 ^ 7 mm) * --- ^ --------------- -Order --------- ^^ 1 '(Please read the notes on the back before filling (This page) 479413 A7 B7 V. Description of the invention (11) (Please read the precautions on the back before filling in this page) The information can be used to greatly reduce the amount of processing. And 'About t = 1 ~ 65 range The scattered data outside is not supposed to be ignored and ignored due to the consideration of processing volume or accuracy, but it is not necessary in theory, so the error caused by cutting off the mantissa will not occur. Figure 3 shows this implementation The overall structure diagram of the D / A converter in the form. The D / A converter shown in FIG. 3 is composed of a digital waveform generation unit 10, a convolution operation unit 20, and a D / A conversion unit 30. The above-mentioned digital waveform generation section 10 corresponds to the synthesis means of the present invention, the convolution operation section 20 corresponds to the supersampling means and operation means of the present invention, and the D / A conversion section 30 corresponds to the D / A conversion means of the present invention. The structure of the digital waveform generation unit 10 can be explained with reference to FIG. 4 as follows. The convolution operation unit 20 is used to perform the oversampling and convolution operations as described with reference to FIG. 6 above to generate digital data 各 for each sampling point among the scattered data of the input digital waveform generation unit 10. The D / A conversion unit 30 is used to perform D / A conversion on the digital data obtained by the convolution operation unit 20 (the conventional interpolation by supersampling is not performed here). The consumer cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs printed the structure of the above-mentioned convolution operation unit 20. The D-type flip-flop (hereinafter referred to as D · FF) la is maintained from the digital waveform generator by a clock pulse (4CK) with twice the frequency. 1 0 digital data output. D · F F lb in parallel with D · F F 1 a also holds the digital data output from the digital waveform generating unit 10 in accordance with the clock pulse (4CK) of twice the frequency. But this side is maintained with the period of phase inversion of 4CK mentioned above. The adder 2 is used to add the digital data held by the two D · F F 1 a and 1 b above. Therefore, D · FF 1 a, 1 b, and adder 2 structure -14- This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 479413 A7 B7 V. Description of the invention (12) Cost invention The first operation means performs two times of oversampling on the digital data output by the digital waveform generating section 10, and adds two samples to each sample thus obtained and each sample staggered by half of its phase. Convolution operation (see Figure 6 (A)). The two D · F F 3a, 3b connected in parallel at the rear stage of the above-mentioned adder 2 will hold the digital data output by the adder 2 and then hold the clock pulses (8CK) with twice the frequency at a period staggered from each other by half. The adder 4 adds the digital data 保持 held by the two D · F F 3a, 3b. Thus, D · FF 3a, 3b and the adder 4 constitute the second operation means of the present invention. The digital data obtained by the convolution operation in the first stage is then subjected to double oversampling. A convolution operation of each sample 段 and two pieces of samples 値 staggered by half of the phase (refer to FIG. 6). The two D · F F 5a and 5b connected in parallel at the rear stage of the above-mentioned adder 4 will hold the digital data output by the adder 4 and then hold the clocks (16CK) with twice the frequency with a period staggered from each other by half. The adder 6 adds the digital data 保持 held by the two D · F F 5a and 5b. In this way, D · FF5a, 5b and the adder 6 constitute the third operation means of the present invention. The digital data obtained by the convolution operation in the second stage is then double-supersampled, and each of the obtained A convolution operation of the sample 値 and the two steps of adding the samples 各 staggered by half of the phase (see FIG. 6 (C)). In this way, three times twice the oversampling and two-stage convolution calculations are performed, and the eight-fold super paper size is applied to the digital data output by the digital waveform generation unit 10. The Chinese national standard (CNS) A4 specification is applicable ( 210 X 297 mm) ---: ------- #-pack --- (Please read the notes on the back before filling this page) · Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs-15- Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs + / 9413 -------- B7_______ V. Description of Invention (13) Sampling. The above-mentioned structure in the convolution operation section 20 corresponds to the supersampling component of the present invention ', and the rest of the structures described below correspond to the operation component of the present invention. The eight D · F F 7a ~ 7h's connected to the subsequent section of the adder 6 are delayed by a clock pulse of 16 times the frequency (16CK), and are sequentially held by one clock pulse. These eight D · F F 7a to 7h correspond to the delay members of the present invention. The rest of the architecture described below corresponds to the addition component of the present invention. The adder 8a and the 1/2 multiplier 9a add the digital data held by D · F F 7g, 7h and multiply by 1/2. The adder 8b and the 1/2 multiplier 9b add the digital data held by D · F F 7e, 7f and multiply by 1/2. The adder 8c and the 1/2 multiplier 9c add the digital data held by D · FF7c, 7d and multiply by 1/2. The adder 8d and the 1/2 multiplier 9d add the digital data held by D · F F 7a, 7b and multiply by 1/2. At the same time, the adder 8e and the 1/2 multiplier 9e will add the digital data output from the two 1/2 multipliers 9a, 9b and multiply by 1/2. The adder 8f and the 1/2 multiplier 9f will be divided by two. The digital data output by the 1/2 multipliers 9c, 9d are added and multiplied by 1/2, the adder 8g and the 1/2 multiplier 9g will be the digital output by the two 1/2 multipliers 9e, 9f. The data is added up, and the result is supplied to the D / A conversion unit 30. With the above D · FF7a ~ 7h, adders 8a ~ 8g, and 1/2 multipliers 9a ~ 9f, the digital data with 16 times oversampling is applied to the clock pulses at 16 times the frequency (1 6 CK) A convolution operation of 8 segments is performed by staggering and adding 1 clock pulse to each sample (refer to FIG. 6 (D)). This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) -T6- (Please read the precautions on the back before filling this page)

479413

V. Description of the invention (1 spoon D / A conversion unit 30 simply performs D / A conversion on each sample of the digital data thus obtained, and continuously outputs smooth signal waveforms of analog data. -------- ------- Install --- (Please read the precautions on the back before filling out this page) Then refer to Figure 4 to explain the structure of the digital waveform generator 10 above. In Figure 4, 3 D · FF 1 1 a ~ 1 1 c Delays the scattered data of the digits of the D / A conversion object by 1 clock pulse according to the clock pulse (CK) of the reference frequency, and sequentially holds them. These 3 D · FF 1 1 a to 1 1 c correspond to the n delay members of the present invention. The -1 times multiplier 12a multiplies the data held by the above-mentioned D · FF 11 a by -1, and the 1 times multiplier 13 a divides the above D · · FF. The data held by 1 1 a is multiplied by 1 (the data at this time remains unchanged). The multiplication results of these multipliers 12a and 13a are based on the clock pulse (CK) of the reference frequency with an action ratio of 1/2 ( The duty ratio) is switched by the switch 14a, and is selectively output to the adder 16. This multiplier 16 is in addition to the multiplication result of the -1 times multiplier 12a or the 1 times multiplier 13a. In addition, the multiplication result of 8 times multiplier 15 is also input, and the two inputs are added to output. The above 8 times multiplier 15 multiplies the data held by D · FF lib by 8 times. Employees of Intellectual Property Bureau, Ministry of Economic Affairs Printed by a consumer cooperative and the -1 times multiplier 1 2a multiplies the data held by D · FF 1 1 c by -1, the 1 times multiplier 1 3b multiplies the data held by D · FF 1 1 c by 1 (at this time The data 値 remain unchanged.) The multiplication results of these multipliers 12b, 13b are switched by the switch 14b according to the clock pulse (CK) of the reference frequency at an action ratio of 1/2, and are selectively output to D · FF 17a. D · FF 17a will be selectively output by switch 14b, the multiplication result of the above -1 times multiplier 12a or 1 times multiplier 13a will be held by the clock pulse (2CK) with twice the frequency. D · FF 17b will be added by ΤΓ output by the device 16 This paper size is in accordance with China National Standard (CNS) A4 (210 X 297 mm). Printed by the Employees ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs. Doubled clock pulse (2CK) is maintained. Adder 18 and D · FF 19 Add the data output from the two D · FF17a, 17b, and hold it with the clock pulse (2CK) with twice the frequency, and then output it to the deduction calculation unit 20 in the lower stage shown in Figure 3. The digital waveform generator 10 configured as described above can process the scattered data of the D / A conversion object and obtain it. The amplitude and amplitude of each scattered data are adjusted according to the basic digital waveform shown in Figure 5. The result of applying a 3-stage convolution operation. As described above, in the present embodiment, when an interpolation 値 is obtained, in the case where there is a rigorous function of the specimen, as long as the scattered data in a range with a finite 値 other than 0 can be considered, this example uses 3 Scattered data for convolution. Fig. 8 shows an example of the operation of the digital waveform generator 10. Fig. 8 (A) shows an example of the scattered data input to the digital waveform generation unit 10 'The horizontal axis represents time, and the vertical axis (a to f) represents the size of the scattered data. Fig. 8 (B) shows a case where the amplitude is adjusted according to the size (a to f) of the dispersed data, and the basic digital waveform shown in Fig. 5 is modulated, and then a convolution operation is applied to these. That is, ′ adds up the data arranged in the vertical direction to output. The result of the digital convolution of the digital waveform generation unit 10 as described above is passed through the convolution operation unit 20 shown in Fig. 3 to obtain each interpolation unit that oversamples the original dispersive data 16 times. The D / A conversion section 30 simply performs D / A conversion on each sample of the digital data thus obtained, and can continuously output, such as the smoothed analog signal waveform of oversampling according to the specimenization function of FIG. 7. As detailed above, in accordance with the present invention, the Chinese National Standard (CNS) A4 specification (210 x 297 mm) is only applied according to the input paper size of the dispersed capital paper. -15 ^ " (Please read the notes on the back before filling (This page)

Printed by the Employees' Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 479413 A7 B7___ V. Description of the invention (16) The digital data corresponding to the basic waveform of the specimenized function is synthesized by convolution operation, and the obtained data is oversampled And convolution operation 'can obtain continuous interpolation. Therefore, unlike the conventional configuration of a sample holding circuit or a low-pass filter, it is possible to suppress the deterioration of the phase characteristics caused by the filter. At the same time, the function 'generated from the basic digital waveform' of this embodiment is a function which converges to 0 at a limited sample position, so the number of scattered data to be considered in order to obtain an interpolation can be limited, which can reduce Processing capacity. Moreover, since the error caused by the trimming mantissa does not occur, an output waveform with little distortion can be obtained. This can greatly improve the quality of analog signals at the output. At the same time, the continuity interpolation required to obtain a smooth analog signal in this embodiment is all obtained through digital processing. Therefore, compared with the traditional analog processing, the processing volume can be much less. There are also benefits for mass production. Then use FIG. 9 to explain the scattered data from the digital basic waveform corresponding to FIG. 5 (-1, 1, 8, 8, 1, -1) / 8, which is generated by η times oversampling and convolution operations. Another processing example of interpolation. In addition, Figure 9 shows an example of 4 times oversampling due to the limitation on the drawing. However, it is also possible to carry out oversampling with a larger magnification (for example, 8 times, 16 times, ...-). In Figure 9, one of the series shown in the leftmost column is a series of 4 times oversampling the original scattered data (-1, 1, 8, 8, 1,-1) / 8. . From the far left to the right, the columns are divided into four columns. The $ paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) -19-~ ---- j ------ -Equipment ---- (Please read the precautions on the back before filling this page) Order i — Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 479413 A7 B7___ 5. Description of the invention (17) Each row is shifted downward by 1 space. The direction of the column in Fig. 9 indicates the time axis. The so-called shifting of the number series downwards corresponds to the delay of the number series shown in the leftmost column. That is, from the left column to the second column, the number column shown in the leftmost column is shifted from the quarter of the clock pulse 4CLK with a frequency of 4 times the frequency. The third column from the left column is staggered from the left column of the second column from the left column by 4 times the frequency of the clock pulse 4CLK. The number column in the fourth column is the number column in which the clock pulse 4CLK with 4 times the frequency is shifted from the number column shown in the third column from the left. At the same time, from the left column to the fifth column, the rows corresponding to the first to fourth columns are added and divided by 4. By performing the processing from the fifth column from the left on this, a four-times super sampling with a four-phase convolution operation is performed digitally. From the 5th column to the 4th column to the right (the 5th to 8th column from the left), the number column shown in the 5th column is shifted downward by 1 space. The 9th column from the left is the sum of the corresponding rows of the 5th to 8th columns and divided by 4. With this processing of the 9th column from the left, it is possible to digitally perform four times of oversampling with four-phase convolution operation. At the same time, from the left to the 10th column, the number shown in the 9th column is shifted downward by 1 space. From the left, the first 11 歹 ij (the rightmost column) number series is the sum of the corresponding rows of the 9th column and the 10th column and divide by 2. The rightmost column is within the purpose. The paper size applies the Chinese National Standard (CNS) A4 (210 X 297 mm) -20-_ (Please read the precautions on the back before filling this page)

479413 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Description of Invention (18) Insertion. Fig. 10 shows a graph obtained by arranging the finally obtained number shown in the rightmost column of Fig. 9. The function with the waveform as shown in Figure 丨 is' differentiable once in the whole field, and when the specimen position t along the horizontal axis is between 丨 and 33, it has a finite value other than 0, and in other fields 领域A function of a finite set of all zeros. At the same time, the function of Fig. 10 is a specimenization function with the feature of taking maximum 値 only at the specimen point at t = 17, and 値 becoming 0 at the four specimen points at t = 1, 9, 25, and 33. All sampling points needed to obtain smooth waveform data. In this way, the function in Fig. 10 is a specimenization function, which can be differentiated once in the whole field, and converges to a finite station of 0 at the specimen position t = 1, 33. Therefore, using the specimenization function of Fig. 10 to perform overlap based on each scattered data, a function that can be differentiated once can be used to interpolate the scattered data. When interpolating, just consider the scattered data in the range of t = 1 ~ 33. Therefore, in order to obtain one interpolation threshold, it is only necessary to consider a limited number of η scattered data, which can greatly reduce the processing amount. Moreover, the scattered data outside the range of t = 1 ~ 33 is not to be ignored due to consideration of the processing volume or accuracy, but is not necessary in theory, so it will not be caused by cutting off the mantissa. error. Therefore, when using the data interpolation method of this embodiment, a correct interpolation can be obtained. Figure 11 is a block diagram of a structure example that does not need to implement the data interpolation processing described in Figure 9 above. The data interpolation device shown in Fig. 11 is composed of a normalized memory unit 2 1. a phase shift unit 2 2. a majority digital multiplier 2 3 a ~ This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 Public)

2T • ---. J -------- install --- (Please read the precautions on the back before filling out this page) 5J · 479413 A7 B7 V. Description of the invention (19) 23d, majority digit addition The devices 24a to 24c and the PLL circuit 25 are configured. The above-mentioned normalized data storage unit 21 is shown in the rightmost column of FIG. 9 and is staggered by 4 phases. The normalized data row is stored (the digital basic waveform is oversampled by 4 times and normalized by the convolution operation. ). When the digital basic waveform shown in FIG. 5 is oversampled by 8 or 16 times, the digital basic waveform stored in the normalization memory 21 is oversampled by 8 or 16 times and convolution is performed. The normalized data rows of the operation are stored in the normalized data of the 4th phase of the normalized data storage section 21, which are read out according to the clock pulses CLK and 8CLK supplied from the PLL circuit 25, and are respectively supplied to four digital multipliers 23a ~ 23d one of the input terminals. The phase shifting unit 22 performs a phase shifting process of shifting the phase of the input scattered data by four phases. The four-phase dispersion data generated by the phase shifting unit 22 is output based on the clock pulses CLK and 8CLK supplied from the PLL circuit 25, and is supplied to the other input terminals of the four digital multipliers 23a to 23d. The four digital multipliers 23a to 23d are multiplied by the normalized data of the four phases output from the normalization registering unit 21 and the scattered data of the four phases output from the phase shifting unit 22. The three digital adders 24a to 24c connected to these subsequent stages add all the multiplication results of the four digital multipliers 23a to 23d and output them. The structure shown in FIG. 11 is the normalized data in the rightmost column obtained from the convolution operation shown in FIG. 9 and stored in the normalized data storage unit 21 such as a ROM in advance. And this normalized data is adjusted to the corresponding volatility of the scattered data input, and then the 4-phase convolution operation is used to add the obtained data to this paper. The paper size applies the national standard (CNS) A4 specification (210 X 297 public). (%) (Please read the notes on the back before filling out this page) Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs

-22- Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 479413 A7 B7_____ V. Description of Invention (2) is output by synthesis. It is also possible to multiply the amplitude 输入 of the input scattered data by the digital basic waveform shown in Figure 5 and perform convolution operations as shown in Figure 9 when interpolating the obtained data 値, but the structure is as shown in Figure 1 In the case of graphs, it is not necessary to perform the convolution operation of FIG. 9 during actual interpolation, which has the advantage that the interpolation process can be speeded up. In addition, the embodiment described above is only an example of a specific embodiment when the present invention is implemented, and the technical scope of the present invention should not be interpreted in this way. That is, the present invention can be implemented in various forms without departing from the spirit or main characteristics thereof. For example, the convolution operation unit 20 shown in Fig. 3 performs double oversampling three times, but the present invention is not limited to this number. At the same time, the convolution operation of 8 segments is performed after the total 8 times of oversampling, but the number of segments is not limited by this. At the same time, the circuit architecture itself for performing such oversampling and convolution operations is not limited to the example shown in FIG. 3. Further, the digital waveform generating unit 10 shown in Fig. 4 performs a convolution operation of three stages, but the present invention does not limit the number of stages. Meanwhile, in the embodiment of FIG. 3 described above, each interpolation 求 obtained by the digital waveform generation unit 10 and the convolution operation unit 20 is finally subjected to D / A conversion by the D / A conversion unit 30. The obtained interpolations do not perform D / A conversion, but are used for other digital processing. That is, a structure in which the D / A conversion section 30 of Fig. 3 is not provided may be used as a data interpolation device. The digital basic waveform in the above embodiment is -1, 1, 8, 8, 1, and -1, but the digital basic waveform is not limited to this example. That is, as long as the interpolation function is differentiable once in the entire field, and converges in a limited number of specimen positions, the paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) ^ 23-" ~~ (Please read the notes on the back before filling this page)

479413 Λ7 B7 V. Description of the invention (21) As a function of a finite station of 0, any waveform is acceptable. For example, the weights of both sides can be 1 or 0 instead of -1. The weight of the middle part can also take any value other than 8. In any case, good curve interpolation can be achieved. At the same time, the D / A conversion unit and the data interpolation device of the present embodiment described above can be configured by hardware as described above, but can also be implemented by a DSP or software. For example, when implemented in software, the device according to this embodiment is actually constituted by a computer's CPU or MPU, RAM, ROM, and the like ', and is realized by a program operation memorized in RAM or ROM. Therefore, a program that can operate a computer to exert the functions of the embodiment described above can be recorded in a CD-ROM medium such as CD-ROM, and read into a computer. In addition to CD-ROM, this recording medium can also use floppy disks, hard disks, magnetic tapes, optical disks, magneto-optical disks, DVDs, non-volatile memory cards, etc. At the same time, it can be realized by downloading the above program to a computer via the Internet or the like. At the same time, not only the computer executes the supplied program to realize the functions of the above-mentioned embodiment. When the program and the OS (operating system) or other application software operating in the computer jointly realize the functions of the above-mentioned embodiment, or the supplied When all or part of the processing of a program is executed by a function expansion board or a function expansion unit of a computer to implement the functions of the above-mentioned embodiments, these programs are still included in the embodiments of the present invention. The present invention suppresses the degradation of the phase characteristics caused by the low-pass filter, and at the same time suppresses the error of the cut-off number during interpolation, thereby suppressing the distortion of the output waveform to a minimum, and greatly improving the analog speech output. The quality of the signal is very useful. This paper size applies to China National Standard (CNS) A4 specification (210 X 297 mm) --- Π -------- install --- (Please read the precautions on the back before filling this page) Order · Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs-24-479413 A7 B7 V. Description of the invention (22) 7: Simple illustration of the diagram The first diagram is an explanatory diagram of the sine function. Fig. 2 is an explanatory diagram of the interpolation operation. Fig. 3 is a diagram showing an example of the structure of the D / A converter according to this embodiment. Fig. 4 is a diagram showing an example of the structure of the digital waveform generating section in Fig. 3. Fig. 5 is a diagram showing a basic digital waveform used in this embodiment. FIG. 6 is a diagram illustrating an operation example of a convolution operation unit. Fig. 7 is a diagram showing a function generated from the basic digital waveform of Fig. 5 by the convolution operation section of Fig. 6; Fig. 8 is a diagram for explaining the operation of the digital waveform generating section in Fig. 3; Fig. 9 is a diagram for explaining other operation examples of the convolution operation section of this embodiment. Fig. 10 is a diagram showing a function generated from the basic digital waveform of Fig. 5 by the convolution operation section of Fig. 9; FIG. 11 is a diagram showing another example of the architecture of the data interpolation device. Main components comparison table la, lb, 3a, 3b, 5a, 5b, 7a ~ 7h, 11a ~ lie, 17a, 17b, 19 ......... D type flip-flop (D · FF ) 2, 4, 6, 8a to 8g, 16, 18 ......... The paper size of the adder is applicable to the Chinese National Standard (CNS) A4 specification (210 x 297 mm). ·- -^ ------- Install --- (Please read the notes on the back before filling this page) • ldl- · Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 479413 A7 B7 V. Invention Description (23 ) 9a ~ 9f---.. 1/2 multiplier 10... Digital waveform generators 12a, 13a ..... .Multipliers 14a, 14b ......... Switch 15 ......-..... 8 times multiplier 20 ......... pleats Product operation unit 21 ......... Normalized data memory unit 22 ......... Phase shift units 23a to 23 < 1 .... ..... digital multipliers 24a ~ 24c ......... digital adder 25 .........- PLL circuit 30 ......... ..D / A Conversion Department ------------ ^^-install --- (please read the precautions on the back before filling this page) The paper size of the paper is applicable to China National Standard (CNS) A4 Specifications (210 X 297 mm) -26-

Claims (1)

479413 A8 B8 C8 D8 6. Scope of patent application (please read the cautions on the back before filling this page) 1. A digital-to-analog converter, characterized by oversampling and more than one moving average calculation or folding The operation is to synthesize the digital data corresponding to the basic waveforms of the input n scattered data, so as to obtain the digital interpolation of the scattered data, and then transform each digital data including the interpolated data. By analogy. 2. —A kind of digital-analog converter, which is characterized in that by oversampling and moving average operation or folding operation, digital data corresponding to the basic waveforms of the input n scattered data are synthesized with each other, The digital data 値 is then subjected to a moving average operation or a folding operation to obtain the interpolated 値 of the digits of the above-mentioned scattered data, and then the digital data 包含 containing the interpolated 値 is converted into an analog quantity. 3. —A kind of digital-analog converter, which is characterized by: having a synthetic component that synthesizes digital data of basic waveforms corresponding to the input n scattered data to each other by moving average operation or folding operation; economical; The Intellectual Property Bureau employee consumer cooperative prints within a few segments, sampling the digital data generated by the above-mentioned synthetic component, twice the frequency of the previous segment, sampling the input data, and obtaining the data, and Add each of these pieces of data out of a certain phase, and output them to the supersampling means for processing in the next paragraph; an arithmetic means for performing a moving average operation or a folding operation on each of the data obtained by the above-mentioned supersampling means; and , Each data obtained by the above-mentioned operation means is transformed into a D / A conversion means of analog quantity. I 4. A digital-to-analog converter, which is characterized by: -27- This paper size applies Chinese National Standard (CNS) A4 specification (210X297 mm) 479413 A8 B8 C8 D8 6. Scope of patent application (please first (Read the precautions on the back and fill in this page again.) The digital data corresponding to the basic waveforms of the η scattered data input according to the reference frequency clock pulses will be separated from each other and added to the reference frequency clock pulses for data synthesis. Synthesizing components; within a few segments, the digital data generated by the above-mentioned synthetic components is sampled from the input data 两倍 twice the frequency of the previous paragraph, each data 値 will be obtained, and these are staggered by half respectively The data of each clock pulse 値 are summed separately and output to the supersampling means of the next stage processing; For each data obtained by the supersampling means 値, the clock pulse at the frequency of the last stage of the supersampling means Next, each data is staggered by one clock pulse and added to perform a moving average or folding operation. And, each of said operation member information obtained is converted into analog Zhi amounts of D / A conversion member. 5. If the digital-to-analog converter in item 4 of the scope of patent application, the special feature is that the synthetic component is provided with: the scattered data input in accordance with the reference frequency clock pulse sequence is sequentially delayed sequentially by one minute η delay components; The Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs prints each data 从 output from the above η delay components, multiplies each gain 对应 corresponding to the basic digital waveform, and adds the multiplication results The multiply-add component that is output to the supersampling component. 6. The digital-to-analog converter according to item 4 of the scope of patent application, characterized in that the above-mentioned supersampling component is provided with: each of the digital data generated by the above-mentioned composite component; Clock pulse sampling, each paper size obtained is applicable to China National Standards (CNS) A4 specifications (210 × 297 mm) I 28-'479413 A8 B8 C8 D8 Patent range data and the first calculation component that adds the data staggered by half a clock pulse; each data obtained from the first calculation component is sampled with a clock pulse that is 4 times the reference frequency A second operation component that adds each of the obtained data and each data that is staggered by half a clock pulse; and, for each data obtained by the second operation component, the Clock pulse sampling at 8 times the frequency, and the third operation means for adding the obtained data and the data separated by half a clock pulse separately.7. For example, the digital-to-analog converter of the third scope of the patent application, characterized in that the computing means is provided with: the digital data obtained by the supersampling means are sequentially delayed in the frequency of the last stage of the supersampling means, respectively. The majority of the delay means of the clock pulses; and an addition means which outputs the above-mentioned majority of the delay means, respectively, and outputs. 8. —A kind of digital-analog conversion method 'characterized in that' contains: by oversampling and more than one moving average operation or folding operation, the digital data corresponding to the basic waveforms of the input n scattered data are synthesized with each other In order to obtain the calculation steps of the interpolated digits of the above-mentioned scattered data; and the D / A conversion step of converting each digital data 包含 including the interpolated 値 obtained by the above-mentioned operations into an analog quantity. 9. A digital-analog conversion method 'characterized by' contains: (Please read the precautions on the reverse side before filling out this page) k-pack ·, 1T. This paper size applies to China National Standard (CNS) Α4 specification (2ΐ〇 χ297 mm) -29- 479413 A8 B8 C8 D8 VI. Patent application range The synthetic steps of synthesizing the digital data of the basic waveforms corresponding to the input n scattered data by using moving average calculation or folding operation; The synthesized digital data 値 is then subjected to an oversampling step accompanied by a moving average operation or a folding operation; for the data 藉 obtained by the above oversampling, a moving average operation or a folding operation is performed to obtain the above-mentioned dispersion. A step of calculating the digital interpolation of the data; and transforming each of the digital data including the interpolation obtained by the above-mentioned operation into an analog D / A conversion step. ίο.Ηϋ 眞 The material interpolation device is characterized in that, by oversampling and averaging operation or folding operation more than-! _ times, it will correspond to the input η points ... 1 basic waveform of scattered data The digital data are combined with each other to obtain the digital interpolation of the above data. 11. The printed material interpolating device for the consumer cooperative of employees of the Intellectual Property Bureau of the Ministry of Economic Affairs is characterized in that, by moving average calculation or folding 输入 #, digital data corresponding to the basic waveforms of the η dispersed data input are combined with each other , Supersampling the synthesized digital data 値, and then performing moving average operation or folding operation on the data 由此 thus obtained, to obtain the interpolation digits of the above-mentioned scattered data. -30- (Please read the notes on the back before filling in this page) The paper size is applicable to the Chinese National Standard (CNS) A # specification (21〇297 mm)